- Problem Overview
- What Data is Used?
- Regression Analysis
- Spatial Analysis
November 20th, 2017
Overview
Problem Overview
Question 1
Does the proximity of opening a new station effect the ridership of other stations near the newly opened station?
- Not able to turn this into a 'spatial data' problem, seems like a distance problem
Question 2
How does the proximity of NYC subway stations effect ridership?
- Bike Share guide recommends placing bike stations near public transit. NACTO
Data Sources
Citibike Data
Source
Observations: 9,884,307 Variables: 11 $ tripduration <dbl> 997, 1904, 305, 250, 464, 1118, 394, 1449, 42... $ starttime <dttm> 2013-07-01 06:00:16, 2013-07-01 06:00:30, 20... $ stoptime <dttm> 2013-07-01 06:16:53, 2013-07-01 06:32:14, 20... $ startstationid <chr> "436", "294", "385", "271", "477", "488", "30... $ startstationname <chr> "Hancock St & Bedford Ave", "Washington Squar... $ endstationid <chr> "467", "375", "440", "390", "522", "497", "32... $ endstationname <chr> "Dean St & 4 Ave", "Mercer St & Bleecker St",... $ bikeid <dbl> 16199, 20281, 18143, 16370, 15497, 15502, 161... $ usertype <chr> "Subscriber", "Subscriber", "Subscriber", "Su... $ birthyear <dbl> 1979, 1949, 1988, 1962, 1975, 1957, 1963, 195... $ gender <chr> "2", "1", "1", "1", "1", "1", "2", "1", "1", ...
Citibike Data
Summary Statistics
Time: 6 - 10 am weekday mornings
Date Range:
[1] "2013-07-01 UTC" "2017-07-31 UTC"
Stations:
[1] 749
Citibike Data
Exploration
Citibike Data
Exploration (Cont.)
Citibike Data
Only the stations labeled in red will be used in the analysis.
These stations have the largest amount of longitudinal data and are more homogenous in the surrounding area.
Start analysis in 2014, giving 6 months "burn in"
Historic Weather Data
NOAA Data Request
Observations: 1,673 Variables: 6 $ day_trip <dttm> 2013-01-01, 2013-01-02, 2013-01-03, 2013-01-04, 2013... $ PRCP <dbl> 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00,... $ SNOW <dbl> 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0... $ SNWD <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,... $ TMIN <int> 26, 22, 24, 30, 32, 34, 37, 35, 39, 40, 37, 42, 43, 3... $ TMAX <int> 40, 33, 32, 37, 42, 46, 45, 48, 49, 47, 46, 47, 50, 5...
Subway Entrance Data
Regression Analysis
Model Definition
- Fit a hierarchical Poisson model to control for variability throughout the year.
- Outcome - Number of Rides
Coefficients:
- Fixed: Temperature, Snow, Days from Jan 1, 2014, previous Days count
- Random: Station, Week of Year
Model Performance
Specific Group of Stations
Spatial Analysis
Models
Does proximity to subway stations show decrease in Citibike usage?
Fitting the Variogram
No Regressors
Closest Subway Features
